548 
Fishery Bulletin 99(4) 
where S = average gut fullness of field-caught fish; and 
R e = exponential evacuation rate at measured 
field temperatures determined from labora- 
tory experiments described above. 
This simplified version of the Elliott and Persson (1978) 
model is appropriate when sampling is not conducted in 
discrete time intervals. We used gut-fullness measures 
from fish collected throughout the day, as described above. 
Although fish were sampled only during daylight hours, 
the sampling interval (24 hours) is substantially shorter 
than the evacuation time (>72 hours), even at the warmest 
temperatures observed; hence any diel feeding patterns 
will have little effect on our consumption estimates. 
Standard deviations of consumption estimates were gen- 
erated from a Taylor expansion of the consumption model 
dC ' 
Z 
2 
a R + 
' dC 1 
Z 
< 4+2 
' dC " 
‘ dC ' 
[dS 
l dR e J 
s 
l dR e ) 
dS ) 
The variance in the evacuation rate parameter (R e ) was 
estimated through a bootstrap procedure by fitting the 
evacuation model to 1000 sets of 152 observations sampled 
with replacement from the evacuation rate data. Since R c 
is a function of temperature, the variance in R e at a given 
temperature was determined by inserting the_tempera- 
ture into the 1000 model fits. The variance in S was esti- 
mated for each date as var(S-)/n. 
The standard deviation of the consumption estimate de- 
pends upon the covariance between evacuation rate (R e ), 
measured in laboratory experiments, and gut fullness ( S ), 
observed in wild fish. Because we have no way of measur- 
ing the covariance between these parameters, we evaluat- 
ed the importance of this term by calculating the variance 
under three assumptions: 1) Sand R e are not correlated, 
2) they covary perfectly, and 3) they display perfect nega- 
tive covariance. 
Analysis of feeding patterns 
We investigated feeding patterns ofYOY striped bass by 
examining the relationship between several factors (body 
size, time of year, water temperature, and energy storage) 
and gut fullness, at the individual and population level. Gut 
fullness was chosen over consumption rate for three rea- 
sons. First, gut fullnesses were measured directly, whereas 
consumption rates were estimated from a model by using 
gut fullnesses. Second, consumption estimates are directly 
dependent on temperature, an independent variable in 
these analyses. Finally, we were interested in determining 
the conditions that stimulate feeding, which we believe are 
more immediately reflected in the gut fullness measures. 
Feeding patterns at the individual level were investi- 
gated by examining the relationships between gut fullness 
and the independent variables of body size, lipid level, wa- 
ter temperature, and time of year. Gut fullness values of 
individual fish were dominated by zeros (empty stomachs) 
and could not be transformed appropriately, preventing 
the use of parametric statistics. Because empty stomachs 
may reflect either a lack of prey availability or reduced 
appetite, we concentrated analyses at the individual level 
on the slope of the 95% quantile of gut fullness regressed 
against the independent variables (Scharf et al., 1998). 
This technique has been used to examine scatter-plots, 
when boundaries of a relationship between two factors are 
of interest, rather than the mean. In our case we were in- 
terested in the factors that stimulate feeding as opposed 
to developing a predictive relationship between indepen- 
dent variables and gut fullness. The 95% quantile line de- 
scribes the relationship between maximum observed gut 
fullness and the independent variables. Mean gut fullness 
of all fish captured on a given date was used to investigate 
feeding patterns at the population level and was compared 
to average lipid level of fish captured on that date, time 
of year, and water temperature, by using Kendall’s coeffi- 
cient of rank correlation (Sokal and Rohlf, 1981). 
Lipid levels were determined for a subsample of fish 
from the diet analysis. This involved determination of the 
percentage of dry body weight comprising non polar lipids 
(those used primarily for energy storage). Details of the 
compositional analysis procedure can be found in Hurst et 
al. (2000) and Schultz and Conover (1997). Compositional 
analysis was performed on 15 to 40 fish from each sam- 
pling date (except 13 December 1995). The relationship 
between lipid level and gut fullness at the individual level 
was examined by using all fish for which both pieces of in- 
formation were available (zz =587), whereas analyses at the 
population level used the average lipid level observed on a 
given date [n= 28). 
Results 
Gastric evacuation rate 
Evacuation rates of overwintering YOY striped bass 
declined greatly with temperature. At mid-winter temper- 
atures, evacuation rates were among the lowest reported 
for any fish species. Time to 50% evacuation ranged from 
31 hours at 11°C to 101 hours at 2°C (Fig. 1). The best 
fit parameters in the evacuation model were 6 0 =-0. 00685, 
6j= 0.126, c 0 = 25.757, and c 1= -1.8033. 
We observed a lag between feeding and a measurable 
loss of material from the stomach. The length of the ob- 
served lag decreased as temperature increased from 18.2 
hours at 2°C to 6.1 hours at 11°C. The exponential mod- 
els used here to describe digestion fitted the experimental 
data as well as, or significantly better than, other common 
models (linear and square root; Bromley 1994). 
The amount of variability among individuals in evacua- 
tion rate increased as temperature decreased, leading to 
a poorer fit of the evacuation models at the lower temper- 
atures (Fig. 1). Deviations from the model increased sig- 
nificantly as temperature decreased (P=0.018; ANOVA of 
absolute value of sample deviations from evacuation mod- 
el). Body size had no effect on evacuation rate among ju- 
venile striped bass. We found no correlation between re- 
sidual values from the evacuation model and fish length 
(r=-0.03 P=0.685). 
